Table 4 The results of the multi-classification.
Deep learning model | Accuracy (%) | Sensitivity (%) | Specificity (%) | Precision (%) | F1 score (%) | |||||
|---|---|---|---|---|---|---|---|---|---|---|
AP | Lateral | AP | Lateral | AP | Lateral | AP | Lateral | AP | Lateral | |
ResNet 50 | 89.74% | 84.44% | 90.87% | 83.36% | 99.5% | 97.89% | 91.0% | 83.5% | 89.25% | 83.35% |
Inception v3 | 85.71% | 81.16% | 84.12% | 79.2% | 95.21% | 93.01% | 84.25% | 79.25% | 84.25% | 79.5% |
DenseNet 201 | 98.51% | 92.42% | 98.75% | 92.77% | 99.5% | 97.69% | 98.75% | 91.75% | 98.75% | 92.0% |
